OAI/Combinatorics/EditApproximation/Geometry/BandRegret.lean
5 declarations, 0 contain sorry, 0 declared as axioms
Regret of the computed rational band mixtures
The optimistic objective is computed in rational arithmetic. The finite optimizer supplies every predicted vector used here; no maximization oracle or unproved entropy-regret hypothesis appears in the conclusion.
- theorem
OAI.EditApproximation.rationalOptimisticHistory_castline 21 - def
OAI.EditApproximation.computedOnlineVectorline 30 - theorem
OAI.EditApproximation.rationalOnlineLinear_objectiveline 36 - theorem
OAI.EditApproximation.computed_finite_block_entropy_regretline 51Source documentation
Fixed-block regret for the actual rational optimizer, with its prescribed
G^80iterations andH^-3numerical loss.theorem computed_finite_block_entropy_regret {ι β : Type*} [Fintype β] [DecidableEq β] (coordinates : Finset ι) (block : ι → β) (vertices : List (ι → ℚ)) (S H : ℕ) (η κ : ℚ) (rangeBound : ℝ) (gain decrement : ℕ → ι → ℚ) (comparator : ι → ℝ) (hvertices : vertices ≠ []) (hH : 4 ≤ H) (hprobability : ∀ p ∈ vertices, (∀ i ∈ coordinates, (0 : ℝ) ≤ p i) ∧ (∑ i ∈ coordinates, (p i : ℝ)) = 1) (hcovered : ∀ i ∈ coordinates, ∃ p ∈ vertices, (0 : ℝ) < p i) (hcQ : comparator ∈ convexHull ℝ (rationalVertexSet vertices)) (hblocks : ∀ b, (coordinates.filter (fun i => block i = b)).Nonempty) (hblockMass : ∀ p ∈ convexHull ℝ (rationalVertexSet vertices), ∀ q ∈ convexHull ℝ (rationalVertexSet vertices), ∀ b, …Mentions: OAI.EditApproximation.rationalVertexSet
- theorem
OAI.EditApproximation.band_computed_entropy_regretline 124Source documentation
Specialization to the actual child/state coordinates and rational fixed child masses of a nonempty band.
theorem band_computed_entropy_regret {β σ : Type*} [Fintype β] [Nonempty β] [DecidableEq β] [DecidableEq σ] (band : Finset (β → σ)) (hband : band.Nonempty) (mass : β → ℚ) (hpos : ∀ child, 0 < mass child) (S H : ℕ) (η κ : ℚ) (rangeBound : ℝ) (gain decrement : ℕ → β × σ → ℚ) {comparison : β → σ} (hc : comparison ∈ band) (hH : 4 ≤ H) (hη : 0 < η) (hκ : 0 ≤ κ) (hκhalf : κ ≤ 1 / 2) (hηrange : (η : ℝ) * rangeBound ≤ (κ : ℝ)) (hdecrementNonneg : ∀ t < S, ∀ i ∈ bandCoordinates band, (0 : ℝ) ≤ decrement t i) (hrange : ∀ t < S, ∀ i ∈ bandCoordinates band, ∀ j ∈ bandCoordinates band, i.1 = j.1 → |(decrement t i : ℝ) - decrement t j| ≤ rangeBound) (hdecrement : ∀ i ∈ bandCoordinates band, ∑ t ∈ range S, (decrement t i : ℝ) ≤ 1) (hgain : ∀ t < S, ∀ i ∈ bandCoordinates band, …Mentions: OAI.EditApproximation.bandCoordinates